AI SDK Kotlin/ai.torad.aisdk/LanguageModel

LanguageModel

Provider-neutral model surface. Per invariant I-4 / I-9, application code never depends on a specific provider — it depends on this interface, and provider-specific differences live in middleware (see LanguageModelMiddleware).

Two operations: generate for one-shot completions, stream for incremental events. Both accept the same LanguageModelCallParams.

Provider implementations live in their own modules:

aisdk-provider-litert (Android, on-device)
aisdk-provider-mlx (iOS, on-device)
aisdk-provider-anthropic / aisdk-provider-openai (cloud, future)

This module ships only ai.torad.aisdk.providers.MockLanguageModel for tests — real providers are out of scope here.

Inheritors

AnthropicMessagesLanguageModel

MockLanguageModel

Properties

modelId

abstract val modelId: String

Stable identifier for telemetry / routing ("litertlm/gemma-3-1b", "openai/gpt-5").

provider

open val provider: String

Provider tag for the model — "openai", "anthropic", "litert-lm", etc. Mirrors v6's LanguageModelV3.provider (per historical parity gap #40). Lets routing layers (a fallback chain, a tier picker) make provider-aware decisions without parsing modelId.

supportedUrls

open val supportedUrls: Map<String, List<String>>

Map of URL regex → media-tag list describing which URL shapes the provider accepts inline (so the agent doesn't have to download + base64-encode them itself). Mirrors v6's LanguageModelV3.supportedUrls. Empty default — most on-device providers don't accept URLs and fall back to base64.

Functions

generate

abstract suspend fun generate(params: LanguageModelCallParams): LanguageModelResult

One-shot completion.

stream

abstract fun stream(params: LanguageModelCallParams): Flow<StreamEvent>

Streaming completion. Cold until collected, then drives one upstream call per collection.

streamResult

open fun streamResult(params: LanguageModelCallParams): LanguageModelStreamResult

Streaming completion plus metadata available before stream collection. Implementations must keep the returned stream cold, and must not flowOn/emit from a foreign context — the collection context is the caller's choice. Providers can override this to expose request bodies and response headers while preserving the v6 doStream result shape.